我們對機器人時代準備不足

Google’s recent announcement that its DeepMind technology had defeated one of the world’s highest-ranked champions at the ancient game of Go is just one example of the many dramatic advances unfolding in the fields of artificial intelligence and robotics. Machines are rapidly taking on ever more challenging cognitive tasks, encroaching on the fundamental capability that sets humans apart as a species: our ability to make complex decisions, to solve problems — and, most importantly, to learn. DeepMind’s feat was especially remarkable not just because the technology ultimately prevailed, but because the system largely trained itself to do so.

谷歌(Google)最近宣佈其DeepMind技術在古老的圍棋比賽中擊敗了世界排名最高的冠軍之一。這只不過是人類在人工智能和機器人領域取得的許多戲劇性進展的一個例子。機器正在迅速承擔起越來越具挑戰性的認知任務，開始形成使人類有別於其他物種的根本能力：我們做出複雜決定的能力、解決問題的能力，以及（最重要的）學習的能力。DeepMind的功績之所以尤其引人矚目，不僅僅是因爲技術終於佔了上風，而且還因爲它基本上是憑藉自我訓練戰勝了對手。

In the coming decades, machine learning is likely to be the primary driving force behind a Cambrian explosion of applications in robotics and software automation. It won’t be long before the tools and building blocks that enable engineers and entrepreneurs to create smart robotic systems will be so advanced and accessible that nearly any opportunity to leverage the technology will be identified and addressed almost immediately. The near-term future is likely to be transformed not by general purpose robots or AI systems but rather a nearly limitless number of specialised applications. Collectively, these systems are likely to span the entire job market and economy, ultimately consuming nearly any kind of work that is on some level routine and predictable.

在今後幾十年裏，機器學習可能是機器人和軟件自動化應用出現“寒武紀大爆發”（Cambrian explosion，化石記錄顯示絕大多數的動物“門”都在距今5.42億年前的寒武紀時期出現，由此得名——譯者注）背後的主要推動力量。不久之後，能讓工程師和企業家們創建智能機器人系統的工具和構造塊將會如此先進和易於獲得，以至於近乎所有能夠利用這種技術的機遇都會被立即發現和抓住。轉變近期未來的，很可能不是一般用途的機器人，而是近乎無限數量的專業應用。總體而言，這些系統可能覆蓋整個就業市場和經濟，最終接手幾乎所有在某種程度上例行和可預見的工作。

Sceptics will be quick to point out that history clearly shows that advancing technology creates new types of work even as it destroys existing occupations. This process will doubtless continue, but it seems unlikely that sufficient opportunities will be created to absorb the workers pushed out of traditional jobs. To take just one example, consider the impact of self-driving cars. Clearly, the jobs of millions of people who drive taxis or delivery vehicles or work for Uber will be at high risk.

懷疑者將很快指出，歷史清楚地表明，先進技術在破壞現有就業機會的同時還會創造新型的就業機會。這種過程無疑將會持續，但機器人技術似乎不太可能創造足夠就業機會吸收那些被擠出傳統崗位的勞動者。這裏只舉一個例子，想想自動駕駛汽車帶來的影響吧。顯而易見的是，駕駛出租車或投遞車輛、或者爲優步(Uber)工作的數以百萬計的人的就業將面臨極高風險。

On the other hand, building a truly robotic car, capable of operating completely without human intervention, remains a substantial challenge. Autonomous car technology relies heavily on highly detailed advanced mapping of the routes to be driven. The problem is handling the unexpected and infrequent challenges that defy that kind of data-driven approach: the fallen tree that blocks the road, the unscheduled construction or any number of other unpredictable situations that might arise.

另一方面，建造真正的、完全不需人類干預就能運行的機器人汽車依然面臨嚴峻挑戰。自動駕駛技術嚴重依賴極爲詳細的駕駛路線圖。問題在於應對背離這種基於數據方式的意外及偶爾出現的挑戰：倒下的樹木擋在路上，計劃外的建築活動或者其他可能出現的許多無法預測的情況。

An obvious solution presents itself: keep people in the loop just to handle those unusual situations. It’s easy to imagine a future where vehicles operate 99 per cent autonomously, but somewhere a control centre contains specially trained people, ready to take over when a car signals that it has encountered something outside the bounds of its normal operating environment. Those controllers, of course, will be engaged in one of those “new” occupations on which we rest our hopes. But how many of those jobs will there be, relative to the number of driving jobs lost?

一個顯而易見的解決辦法應運而生：讓人留在環路中，以便處理那些異常情況。不難想象未來的車輛在99%的情況下自動駕駛，但在控制中心會有經過特殊培訓的專業人員，他們隨時準備在汽車發出信號表明其遭遇正常運行環境以外的情況時接手。當然，那些控制人員將從事我們寄予厚望的“新”職業之一。但是相比失去的那麼多駕駛工作，會有多少那樣的工作機會？

Needless to say, this mismatch between job destruction and creation isn’t going to be confined to driving. This basic approach — automating nearly all routine and predictable aspects of an occupation and then consolidating the remaining unpredictable tasks into a small number of jobs — is likely to be applied across the board. The low-wage service sector jobs in areas such as fast food and retail, which constitute a substantial fraction of the jobs being created by the economy in both the US and the UK, are certain to be heavily affected. Even more important will be all the white-collar occupations that involve relatively routine information analysis and manipulation. As these “good” jobs, often held by university graduates, begin to evaporate, faith in evermore education and training as the common solution to technological disruption of the job market seems likely to also erode.

不用說，這種就業破壞和創造之間的不匹配不僅侷限於駕駛。這種基本套路——將一份工作的幾乎所有例行和可預見的部分都自動化，然後將剩餘的不可預測的任務整合爲少數的工作崗位——很可能被應用於各行各業。快餐和零售等低薪服務行業的就業機會無疑會受到巨大影響——目前美國和英國經濟創造的就業崗位中有一大部分是在這些服務行業。甚至更爲重要的將是，所有那些涉及相對例行的信息分析和操縱的白領職業都會受到影響。隨着這些往往由大學畢業生從事的“好”工作開始消失，人們很可能不再相信越來越多的教育和培訓是針對技術對就業市場破壞的良方。

All of this portends a social, economic and political disruption for which we are completely unprepared. Widespread unemployment (or even underemployment) has clear potential to rend the fabric of society. Beyond that, it also carries substantial economic risks: in a world with far too few jobs, who will have the income and confidence to purchase the products and services produced by the economy? Where will demand come from? For years, average households in the US have been relying ever more on debt to support their consumption. How will they continue to service those debts in a future where jobs are beginning to evaporate en masse?

所有這些預示着一場我們毫無防備的社會、經濟和政治混亂。普遍失業（甚至不充分就業）顯然有可能撕裂社會架構。此外，它還帶有巨大的經濟風險：在一個就業崗位實在太少的世界裏，誰會有收入和信心購買經濟體生產的產品和服務？需求將會來自哪裏？多年來，美國普通家庭越來越依賴債務支持他們的消費。在就業崗位開始大規模消失的未來，他們如何才能繼續償還這些債務？

In recent years, prominent individuals such as Stephen Hawking and Elon Musk have warned of the risks associated with “killer robots” or super-intelligent machines. While these concerns may some day be relevant, and while there are certainly important ethical considerations involving the use of autonomous systems in military and security applications, I would argue that the most important immediate challenge we face will be adjusting to the economic and social implications of a robotic revolution in the workplace. That disruption is already beginning to unfold, and one might reasonably argue that its impact can already be measured in terms of the political upheaval occurring in both the US and Europe. If we fail to have a meaningful public conversation about what robotics and artificial intelligence mean for the future, and develop workable ways in which to adapt our economy and society, then far greater, and more frightening, volatility is sure to soon arrive.

最近幾年，史蒂芬•霍金(Stephen Hawking)和埃隆•馬斯克(Elon Musk)等知名人士警告了與“機器人殺手”或超智能機器相關的風險。儘管這些擔憂有朝一日會變得相關，儘管在軍事和安全應用場合採用自動化系統確實有重要的倫理課題，但我仍會主張，我們面臨的最重要最緊迫的挑戰將是適應職場機器人革命的經濟和社會影響。這種影響已經開始顯現，人們可以合理地辯稱，從美國和歐洲的政治動盪已經可以看出這種影響。如果我們不能圍繞機器人和人工智能對未來意味着什麼展開有意義的公共討論，並找到讓我們的經濟和社會適應的可行方法，那麼更嚴重更可怕的動盪必定會很快到來。